Available for freelance workSoftware Engineer · Full-Stack Developer
I build software end to end — from the data model to the last pixel.
fabian@buildbyfabi ~ $sudo make me a sandwich
Okay.
fabian@buildbyfabi ~ $rm -rf node_modules && npm i
Ritual complete. The bug is still there.
fabian@buildbyfabi ~ $git log --oneline -1
a1b2c3d fix: it works now, for real this time
fabian@buildbyfabi ~ $:wq
You exited vim on the first try. Nobody will believe you.
fabian@buildbyfabi ~ $echo $STATUS
● Available for freelance work
fabian@buildbyfabi ~ $
About
I am a software engineer, graduated from CUJAE in July 2026. My thesis was a web digital twin to manage, forecast and monitor a photovoltaic microgrid: instead of solving it for a single installation, I designed it around a configurable rule set so it could integrate with different APIs and adapt to very different real-world scenarios.
For the past year I have been working at SunCar as a developer and one of the team leads, building both their corporate site and the internal system the company runs on every day. I also work as a remote freelancer: I am drawn to projects where I can follow the product from start to finish, or join one already in motion and push exactly where the team needs momentum.
My ground is React and Next.js on the frontend, FastAPI and MongoDB on the backend, and Claude Code as a natural part of the workflow — I believe AI-assisted development, used well, raises the bar of the work rather than replacing it. Beyond the stack, what I enjoy most is gathering requirements with the client, modelling the architecture before writing the first line, breaking my own implementations to find where they fail, and building security in from the design stage. And I work better with company: I would rather debate a technical decision with the team than get it right alone.
Stack
What I use on real projects — not a list of everything I've ever touched.
Work
Production systems, my own products, and academic work.
Corporate site and internal system for a solar energy company
I build the software at SunCar, a solar panel installation company. I developed their corporate site in Next.js and lead part of the internal management system: crews and services, material control with kardex, cost sheets and accounting, with PDF and Excel reports and offline support for fieldwork.
Thesis project — solar microgrid monitoring and forecasting
Web platform to manage, forecast and monitor the behaviour of a photovoltaic microgrid. I designed it generically, around a configurable rule set, so it can integrate with different APIs and adapt to any real deployment scenario instead of being tied to a single installation.
AI-assisted shopping platform
Smart shopping platform for restaurants, clothing and groceries, with AI-assisted recommendations. The site is built with Astro and Svelte islands; the project also includes a native Android app in Kotlin for the merchant side.
Booking SaaS for barbershops
Online booking platform for barbershops: each business gets a public page with its own brand, services and opening hours, while the barber manages appointments, staff and gallery from a dashboard. My own product, with an open demo and in active development.
Parcel service from the United States to Havana
Site and dashboard for ShellBox, an agency that buys from US stores and ships the order to Cuba. Astro with server rendering and Svelte islands only where interactivity is needed: a stock catalogue backed by MongoDB, an admin panel with JWT sessions and photo uploads to S3, and every rate centralised in a single configuration file.
Calibrated football forecasting with a reproducible backtest
Probability model for the five big European leagues. The first version trained on statistics from the very match it was predicting, so I rebuilt it from scratch: a single feature definition shared by training and production, a walk-forward backtest over 12,553 matches the model never saw, and a hash-chained ledger that signs every forecast before kick-off.
Self-organizing maps over any CSV
Desktop application that trains self-organizing maps over any numeric CSV, with three topologies, a U-matrix, component planes and automatic hyperparameter search. It started as a team project at CUJAE and I picked it up to decouple it from the dataset it was born with, measure its quality with cross-validation and ship it packaged, with 100 tests and continuous integration.
Contact
Tell me your idea or what you need to solve, and let's work out how to move it forward.
↳Havana, Cuba